FIG. 1 shows an idealized CMOS layout for a transistor 101A. The transistor 101A includes a doped diffusion region 103, which is either p-type or n-type to form source and drain regions of the transistor 101A. The transistor 101A also includes a poly silicon layer used for a gate electrode 101. Additionally, an insulator 105 is deposited to separate the gate electrode 101 from the diffusion region 103. In general, semiconductor device designers control the characteristics of the transistor 101A by varying its dimensions. The dimensions of the transistor 101A are defined by a channel width (W) and a channel length (L). The channel width (W) is defined by the size of the diffusion region 103. The channel length (L) is defined by a size of the gate electrode 101 in the direction extending between the source and drain regions. The channel width (W) is varied to optimize the drive current of the transistor 101A. The channel length (L) is typically defined as small as possible within the limitations of the manufacturing process.
FIG. 2 shows two exemplary transistors 203A and 205A which have different channel widths, and which share a common diffusion node 201A. Transistor 203A is formed by an intersection of its gate electrode 203 and a diffusion region 201. Transistor 205A is formed by an intersection of its gate electrode 205 and the diffusion region 201. The common diffusion node 201A is coupled to a higher level interconnect conductor 207 (e.g., metal 1 line) by way of contact 209. It should be understood that the sharp-cornered periphery of the diffusion region 201 is shown in an idealized manner. In reality, an ability to resolve corners of layout shapes, such as those associated with diffusion region 201, decreases dramatically as the resolution of the lithography process decreases.
FIG. 3 shows a more realistic representation of the as-fabricated diffusion region 201. As shown in FIG. 3, limitations on the resolution of the lithography process causes the corners of the as-fabricated diffusion region 201 to be rounded. For example, at 45 and 32 nanometer (nm) process nodes, sharp corners such as those of diffusion region 201 are well outside of the resolution capability of conventional lithography systems. It should be appreciated that lithography-induced curvature in a diffusion region boundary that extends beneath a gate electrode can cause unacceptable variability in a channel width of the transistor formed by the gate electrode. And, such unacceptable variability in the channel width of the transistor may cause an unacceptable performance of the transistor. An extent of transistor channel width variability due to lithography-induced curvature at diffusion region bend locations depends upon the proximity of the bend locations to the gate electrode, and any misalignment between the gate electrode mask and the diffusion layer, and possibly other processing effects. It is also worth noting that shallow trench isolation (STI) strain, strain layer variation, and other proximity and shape related effects are also dependent upon the as-fabricated diffusion region shape.
One approach for mitigating the effects of lithography-induced curvature in diffusion region boundaries is to separate transistors that share a common node, such that each bend in the diffusion region boundary associated with a given transistor can be moved sufficiently far away from the gate electrode of the given transistor so as to maintain a substantially straight as-fabricated diffusion region boundary beneath the gate electrode of the given transistor. FIG. 4 shows an example application of this approach to define transistors 203A and 205A of FIG. 3. The transistors 203A and 205A are separated and are defined by respective diffusion regions 401 and 403. The separation of transistors 203A and 205A allows the as-fabricated boundaries of diffusion regions 401 and 403 to be straight at the locations beneath gate electrodes 203 and 205, respectively. The shared diffusion node 201A is connected by way of contacts 405 and 407 and higher-level interconnect conductor 409.
It should be understood, however, that separation of transistors that share a common node can be non-optimal. For example, separation of transistors 203A and 205A as shown FIG. 4 results in an increased usage of chip area to define the transistors 203A and 205A. In modern semiconductor devices, chip area is at a premium and such increased chip area usage is undesirable. Also, separation of transistors 203A and 205A as shown FIG. 4 results in an increased capacitance on the shared diffusion node 201A, which can negatively impact the performance of transistors 203A and 205A.
Another approach for mitigating the effects of lithography-induced curvature in diffusion region boundaries requires that transistors be forced to have a same channel width, even when not required for adequate and/or optimal circuit performance. FIG. 5 shows an example application of this approach to define the transistors 203A and 205A of FIG. 3. As shown in FIG. 5, the diffusion region 501 is now defined such that both of transistors 203A and 205A have a common channel width. However, forcing transistors to a common channel width when neither necessary nor optimal can be detrimental to circuit performance. For example, increasing the channel width of transistor 205A over its width as required to meet design constraints causes an increase in leakage and active power consumption of the circuit. Moreover, increasing the channel width of transistors beyond what is necessary can have a negative impact on chip area utilization, because larger drivers may then be required for the oversized transistors.